Using experiential optimization to build lexical representations.

To account for natural variability in cognitive processing, it is standard practice to optimize a model's parameters by fitting it to behavioral data. Although most language-related theories acknowledge a large role for experience in language processing, variability reflecting that knowledge is usually ignored when evaluating a model's fit to representative data. We fit language-based behavioral data using experiential optimization, a method that optimizes the materials that a model is given while retaining the learning and processing mechanisms of standard practice. Rather than using default materials, experiential optimization selects the optimal linguistic sources to create a memory representation that maximizes task performance. We demonstrate performance on multiple benchmark tasks by optimizing the experience on which a model's representation is based.

Release from PI: An analysis and a model.

Recall decreases across a series of subspan immediate-recall trials but rebounds if the semantic category of the words is changed, an example of release from proactive interference (RPI). The size of the rebound depends on the semantic categories used and ranges from 0% to 95%. We used a corpus of novels to create vectors representing the meaning of about 40,000 words using the BEAGLE algorithm. The distance between categories and spread within categories jointly predicted the size of the RPI. We used a holographic model for recall equipped with a lexicon of BEAGLE vectors representing the meaning of words. The model captured RPI using a hologram as an interface to bridge information from episodic and semantic memory; it is the first account of RPI to capture release at the level of individual words in categorized lists.

Implicit learning is order dependent.

We report two experiments using the artificial-grammar task that demonstrate order dependence in implicit learning. Studying grammatical training strings in different orders did not affect participants' discrimination of grammatical from ungrammatical test strings, but it did affect their judgments about specific test strings. Current accounts of learning in the artificial-grammar task focus on category-level discrimination and largely ignore item-level discrimination. Hence, the results highlight the importance of moving theory from a category- to an item-level of analysis and point to a new way to evaluate and to refine accounts of implicit learning.

A computational account of the production effect: Still playing twenty questions with nature.

People remember words that they read aloud better than words that they read silently, a result known as the production effect. The standing explanation for the production effect is that producing a word renders it distinctive in memory and, thus, memorable at test. By 1 key account, distinctiveness is defined in terms of sensory feedback. We formalize the sensory-feedback account using MINERVA 2, a standard model of memory. The model accommodates the basic result in recognition as well as the fact that the mixed-list production effect is larger than its pure-list counterpart, that the production effect is robust to forgetting, and that the production and generation effects have additive influences on performance. A final simulation addresses the strength-based account and suggests that it will be more difficult to distinguish a strength-based versus distinctiveness-based explanation than is typically thought. We conclude that the production effect is consistent with existing theory and discuss our analysis in relation to Alan Newell's (1973) classic criticism of psychology and call for an analysis of psychological principles instead of laboratory phenomena. (PsycINFO Database Record

Applying an exemplar model to an implicit rule-learning task: Implicit learning of semantic structure.

Studies of implicit learning often examine peoples' sensitivity to sequential structure. Computational accounts have evolved to reflect this bias. An experiment conducted by Neil and Higham [Neil, G. J., & Higham, P. A.(2012). Implicit learning of conjunctive rule sets: An alternative to artificial grammars. Consciousness and Cognition, 21, 1393-1400] points to limitations in the sequential approach. In the experiment, participants studied words selected according to a conjunctive rule. At test, participants discriminated rule-consistent from rule-violating words but could not verbalize the rule. Although the data elude explanation by sequential models, an exemplar model of implicit learning can explain them. To make the case, we simulate the full pattern of results by incorporating vector representations for the words used in the experiment, derived from the large-scale semantic space models LSA and BEAGLE, into an exemplar model of memory, MINERVA 2. We show that basic memory processes in a classic model of memory capture implicit learning of non-sequential rules, provided that stimuli are appropriately represented.

Information theory and artificial grammar learning: inferring grammaticality from redundancy.

In artificial grammar learning experiments, participants study strings of letters constructed using a grammar and then sort novel grammatical test exemplars from novel ungrammatical ones. The ability to distinguish grammatical from ungrammatical strings is often taken as evidence that the participants have induced the rules of the grammar. We show that judgements of grammaticality are predicted by the local redundancy of the test strings, not by grammaticality itself. The prediction holds in a transfer test in which test strings involve different letters than the training strings. Local redundancy is usually confounded with grammaticality in stimuli widely used in the literature. The confounding explains why the ability to distinguish grammatical from ungrammatical strings has popularized the idea that participants have induced the rules of the grammar, when they have not. We discuss the judgement of grammaticality task in terms of attribute substitution and pattern goodness. When asked to judge grammaticality (an inaccessible attribute), participants answer an easier question about pattern goodness (an accessible attribute).

Memory as a hologram: an analysis of learning and recall.

We present a holographic theory of human memory. According to the theory, a subject's vocabulary resides in a dynamic distributed representation-a hologram. Studying or recalling a word alters both the existing representation of that word in the hologram and all words associated with it. Recall is always prompted by a recall cue (either a start instruction or the word just recalled). Order of report is a joint function of the item and associative information residing in the hologram at the time the report is made. We apply the model to archival data involving simple free recall, learning in multitrial free recall, simple serial recall, and learning in multitrial serial recall. The model captures accuracy and order of report in both free and serial recall. It also captures learning and subjective organisation in multitrial free recall. We offer the model as an alternative to the short- and long-term account of memory postulated in the modal model.

Editorial.

Mewhort talks about being editor of Canadian Journal of Experimental Psychology. (PsycINFO Database Record (c) 2013 APA, all rights reserved).

Encoding structure in holographic reduced representations.

Vector Symbolic Architectures (VSAs) such as Holographic Reduced Representations (HRRs) are computational associative memories used by cognitive psychologists to model behavioural and neurological aspects of human memory. We present a novel analysis of the mathematics of VSAs and a novel technique for representing data in HRRs. Encoding and decoding in VSAs can be characterised by Latin squares. Successful encoding requires the structure of the data to be orthogonal to the structure of the Latin squares. However, HRRs can successfully encode vectors of locally structured data if vectors are shuffled. Shuffling results are illustrated using images but are applicable to any nonrandom data. The ability to use locally structured vectors provides a technique for detailed modelling of stimuli in HRR models.

Grammaticality is inferred from global similarity: A reply to Kinder (2010).

Jamieson and Mewhort (2009b) proposed an account of performance in the artificial-grammar judgement-of-grammaticality task based on Hintzman's (1986) model of retrieval, Minerva 2. In the account, each letter is represented by a unique vector of random elements, and each exemplar is represented by concatenating its constituent letter vectors. Although successful in simulating several experiments, Kinder (2010) showed that the model fails for three selected experiments. We track the model's failure to a constraint introduced by concatenating letter vectors to construct the exemplar representation. To fix the problem, we use a holographic representation. Holographic representation not only provides the flexibility missing with the concatenation scheme but also acknowledges variability in what subjects notice when they inspect training exemplars. Armed with holographic representations, we show that the model successfully captures the three problematic data sets. We argue for retrospective accounts, like the present one, that acknowledge subjects' skill in drawing unexpected inferences based on memory of studied items against prospective accounts that require subjects to learn statistical regularities in the training set in anticipation of an undefined classification test.

Contrast and assimilation in categorization and exemplar production.

We studied contrast and assimilation in three tasks: an exemplar-production task, a categorization task, and a combined categorization-then-production task. On each trial of the first task, subjects produced a circle when prompted with a category label. In the second task, they classified lines that differed in length into one of four categories. On each trial of the combined task, they classified two lines and then produced a line when prompted by a category label. All three tasks converged on the same conclusion: subjects' representation of the categories (measured in pixels in the production tasks and by the direction of errors in classification) shifted systematically from trial to trial. When successive stimuli were from the same category, the representation of that category was pulled toward the exemplar from the previous trial. When successive stimuli were from different categories, the representations of the neighbouring categories were pushed from the category of the initial stimulus. We conclude that accounts of categorization and identification must accommodate both assimilation and contrast as a function of trial-to-trial shifts in representation.

Global similarity predicts dissociation of classification and recognition: evidence questioning the implicit-explicit learning distinction in amnesia.

Dissociation of classification and recognition in amnesia is widely taken to imply 2 functional systems: an implicit procedural-learning system that is spared in amnesia and an explicit episodic-learning system that is compromised. We argue that both tasks reflect the global similarity of probes to memory. In classification, subjects sort unstudied grammatical exemplars from lures, whereas in recognition, they sort studied grammatical exemplars from lures. Hence, global similarity is necessarily greater in recognition than in classification. Moreover, a grammatical exemplar's similarity to studied exemplars is a nonlinear function of the integrity of the data in memory. Assuming that data integrity is better for control subjects than for subjects with amnesia, the nonlinear relation combined with the advantage for recognition over classification predicts the dissociation of recognition and classification. To illustrate the dissociation of recognition and classification in healthy undergraduates, we manipulated study time to vary the integrity of the data in memory and brought the dissociation under experimental control. We argue that the dissociation reflects a general cost in memory rather than a selective impairment of separate procedural and episodic systems.

Applying the permutation test to factorial designs.

The permutation test follows directly from the procedure in a comparative experiment, does not depend on a known distribution for error, and is sometimes more sensitive to real effects than are the corresponding parametric tests. Despite its advantages, the permutation test is seldom (if ever) applied to factorial designs because of the computational load that they impose. We propose two methods to limit the computation load. We show, first, that orthogonal contrasts limit the computational load and, second, that when combined with Gill's (2007) algorithm, the factorial permutation test is both practical and efficient. For within-subjects designs, the factorial permutation test is equivalent to an ANOVA when the latter's assumptions have been met. For between-subjects designs, the factorial test is conservative. Code to execute the routines described in this article may be downloaded from http://brm.psychonomic-journals.org/content/supplemental.

Applying an exemplar model to the artificial-grammar task: String completion and performance on individual items.

Jamieson and Mewhort (2009a) demonstrated that performance in the artificial-grammar task could be understood using an exemplar model of memory. We reinforce the position by testing the model against data for individual test items both in a standard artificial-grammar experiment and in a string-completion variant of the standard procedure. We argue that retrieval is sensitive to structure in memory. The work ties performance in the artificial-grammar task to principles of explicit memory.

Test sequence priming in recognition memory.

The authors examined priming within the test sequence in 3 recognition memory experiments. A probe primed its successor whenever both probes shared a feature with the same studied item (interjacent priming), indicating that the study item like the probe is central to the decision. Interjacent priming occurred even when the 2 probes did not themselves share any features: A lure that shared a single feature with a study item primed a lure that shared a different feature with the same study item. The experiments distinguished interjacent priming from other types of facilitation. Interjacent priming indicates that a study item that is like the probe is more relevant to the decision than other study items, contrary to global memory models. It also shows that negative decisions depend on contradiction, not insufficient familiarity, because lures, as well as targets, benefited. The data are discussed in terms of a recall check within a dual-process theory, but the authors prefer a single-process resonance model with separate decision mechanisms for yes and no responses (D. J. K. Mewhort & E. E. Johns, 2005).

Randomization tests and the unequal-N/unequal-variance problem.

When both the variance and the N are unequal in a two-group design, the probability of a Type I error shifts from the nominal 5% error rate. The probability is too liberal when the small cell has the larger variance and too conservative when the large cell has the larger variance. We present an algorithm to circumvent the problem when the smaller group has the larger variance and show, by simulation, that the algorithm brings the error rate back to the nominal value without sacrificing the ability to detect true effects.

Applying an exemplar model to the serial reaction-time task: anticipating from experience.

We present a serial reaction time (SRT) task in which participants identified the location of a target by pressing a key mapped to the location. The location of successive targets was determined by the rules of a grammar, and we varied the redundancy of the grammar. Increasing both practice and the redundancy of the grammar reduced response time, but the participants were unable to describe the grammar. Such results are usually discussed as examples of implicit learning. Instead, we treat performance in terms of retrieval from a multitrace memory. In our account, after each trial, participants store a trace comprising the current stimulus, the response associated with it, and the context provided by the immediately preceding response. When a target is presented, it is used as a prompt to retrieve the response mapped to it. As participants practise the task, the redundancy of the series helps point to the correct response and, thereby, speeds retrieval of the response. The model captured performance in the experiment and in classic SRT studies from the literature. Its success shows that the SRT task can be understood in terms of retrieval from memory without implying implicit learning.

Applying an exemplar model to the artificial-grammar task: inferring grammaticality from similarity.

We present three artificial-grammar experiments. The first used position constraints, and the second used sequential constraints. The third varied both the amount of training and the degree of sequential constraint. Increasing both the amount of training and the redundancy of the grammar benefited participants' ability to infer grammatical status; nevertheless, they were unable to describe the grammar. We applied a multitrace model of memory to the task. The model used a global measure of similarity to assess the grammatical status of the probe and captured performance both in our experiments and in three classic studies from the literature. The model shows that retrieval is sensitive to structure in memory, even when individual exemplars are encoded sparsely. The work ties an understanding of performance in the artificial-grammar task to the principles used to understand performance in episodic-memory tasks.